Intelligent respirator operation monitoring system

Abstract

The invention discloses an intelligent respirator operation monitoring system. Comprising a data acquisition module, a safety margin evaluation function definition module, a safety margin standardization module, an initial dangerous point calibration module, a risk direction deviation detection module, a self-adaptive step length updating module, an iterative convergence judgment module, a system training module and a respirator operation monitoring module. The invention belongs to the field of operation monitoring, and particularly relates to an intelligent respirator operation monitoring system, which is characterized in that gradient is used as guidance, continuous iteration approximation to a dangerous point is realized, and a risk peak value is accurately positioned in a respiratory cycle; an included angle between a real risk trajectory and an ideal trajectory is quantified through risk direction deviation detection, and a transient nonlinear event of sudden airway blockage is captured; angle normalization curve amplification is carried out in a breathing stable interval, and approaching to the most dangerous point is accelerated; linear convergence correction is carried out in the nonlinear sudden change interval, false alarm caused by short-term noise is avoided, and the operation monitoring effect is further improved.

Description

Technical Field

[0001] The present invention relates to the field of operation monitoring, and specifically refers to an intelligent ventilator operation monitoring system. Background Art

[0002] A ventilator operation monitoring system is an intelligent platform integrating sensing, data processing, reliability analysis and real-time warning. However, the general ventilator operation monitoring system has the problems of easily missing the transient extreme points in the cycle, not being timely enough in responding to the sudden risks of instantaneous airway obstruction and sudden change of patient compliance, generating false oscillation alarms, and thus resulting in poor accuracy of operation monitoring; the general ventilator operation monitoring system has the problems of slow mutation response, repeatedly oscillating in the non-linear local interval, missing the real change of risk direction, being prone to false alarms and missed alarms, and having poor operation monitoring effect. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent ventilator operation monitoring system. Aiming at the problems that the general ventilator operation monitoring system is prone to miss the transient extreme points in the cycle, not be timely enough in responding to the sudden risks of instantaneous airway obstruction and sudden change of patient compliance, generate false oscillation alarms, and thus result in poor accuracy of operation monitoring, this solution is guided by the gradient, continuously iterates to approach the dangerous point, and accurately locates the risk peak within the respiratory cycle; quantifies the angle between the real risk trajectory and the ideal trajectory through risk direction deviation detection, adjusts the step size in real time, and captures the transient non-linear events of sudden airway obstruction; thereby improving the accuracy of operation monitoring; aiming at the problems that the general ventilator operation monitoring system has slow mutation response, repeatedly oscillates in the non-linear local interval, misses the real change of risk direction, is prone to false alarms and missed alarms, and has poor operation monitoring effect, this solution amplifies the angle normalization curve in the stable breathing interval to accelerate the approach to the most dangerous point; performs linear convergence correction in the non-linear mutation interval, resists oscillation divergence during severe non-linearity, and prevents missed alarms or false alarms; avoids false alarms caused by short-term noise based on double convergence determination, and thus improves the operation monitoring effect.

[0004] The technical solution adopted by the present invention is as follows: An intelligent ventilator operation monitoring system provided by the present invention includes a data acquisition module, a safety margin evaluation function definition module, a safety margin standardization module, an initial dangerous point calibration module, a risk direction deviation detection module, an adaptive step size update module, an iterative convergence determination module, a system training module and a ventilator operation monitoring module;

[0005] The data acquisition module acquires the historical operation data of the ventilator and constructs an operation monitoring sample set;

[0006] The safety margin evaluation function definition module obtains the safety margin evaluation function by calculating the standardized margin of the deviation between each operation monitoring variable and its safety boundary;

[0007] The safety margin standardization module uses the empirical distribution function to map the original operation monitoring variables to the standard normal space;

[0008] The initial dangerous point calibration module uses the most dangerous point within a period as the starting point for the first iterative position update;

[0009] The risk direction deviation detection module analyzes and detects the direction of the risk deviation;

[0010] The adaptive step size update module adjusts the step size based on interval division;

[0011] The iterative convergence determination module performs iterative determination on the operation data within a period;

[0012] The system training module trains the parameters within the system;

[0013] The ventilator operation monitoring module monitors the real-time collected operation data of the ventilator within a period by connecting each system module in series.

[0014] Furthermore, the data acquisition module acquires the historical operation data of the ventilator; labels the operation status of the operation data of the ventilator within a period, and the operation status includes normal operation and abnormal operation; preprocesses the historical operation data of the ventilator to obtain an operation monitoring sample set; sets safety boundaries for each parameter.

[0015] Furthermore, the safety margin evaluation function definition module defines the safety margin evaluation function , to ensure the simultaneous safety of all indicators, the worst margin is taken. For the i operation monitoring samples x i , the safety margin evaluation function is expressed as: ; where is the safety boundary of the jth monitoring variable; is the jth monitoring variable; is the standard deviation of the jth monitoring variable; is the weight of the jth monitoring variable; corresponds to the most dangerous dimension value within the sample; for the operation monitoring sample set, calculate the of each operation monitoring sample, and take the sample corresponding to the minimum value as the initial most dangerous point within a period.

[0016] Furthermore, the safety margin standardization module performs a normalization transformation on each variable, expressed as: ; where is the inverse standard normal CDF; is the empirical distribution function of the i-th monitored quantity; obtain the mapped vector u; rewrite the safety margin evaluation function as: ; the vector u in each mapping space is mapped to the operation monitoring sample .

[0017] Furthermore, the initial dangerous point calibration module in the standard normal space determines the initial most dangerous point within the respiratory cycle as , which serves as the starting point for the iteration of this cycle, and the initial step size , and the position update is: ; where is the position of the most dangerous point after the first iteration; is the value of the safety margin evaluation function at ; is the gradient.

[0018] Furthermore, the risk direction deviation detection module calculates for the k-th iteration: ; ; ; where and are the positions of the most dangerous points at the k-th iteration and the k-1-th iteration respectively; is the value of the safety margin evaluation function at ; is the ideal update vector; is the alignment degree; is the angle deviation.

[0019] Furthermore, the adaptive step size update module sets the turning angle threshold , with a value range of (0, 90) degrees. For the stationary interval of , the step size is adjusted through a curve, expressed as: ; ; where t is the normalized angle deviation; is the step size; for the non-linear mutation interval of , the step size is corrected through linear convergence, expressed as: ; ; ; ; where is the mutation auxiliary function; c is the balance coefficient; is the direction vector; is the smoothing term; take the minimum m such that ; ; where is the control parameter, with a value range of (0, 1); T is the transpose operation, and here the transpose of the gradient vector is taken.

[0020] Further, for the stationary interval, the iterative convergence determination module updates and adopts ; Nonlinear mutation interval update: ; Convergence determination, expressed as: ; Calculate the reliability index , expressed as: ; ; Wherein, is the position of the most dangerous point at the (k + 1)-th iteration; is the momentum coefficient, and its value range is [0, 1]; and are the convergence thresholds, and their value ranges are [10 -3 , 10 -6 ; A reliability index is obtained for each breathing cycle to quantify the safety margin; When , it is determined that the operation within the breathing cycle is abnormal; Otherwise, it is determined to be operating normally; is the alarm probability threshold, and its value range is (0, 1); Cumulative distribution function of the standard normal distribution.

[0021] Further, the system training module divides the operation monitoring sample set into a test set and a training set, trains the system parameters based on the training set, uses the cross-entropy loss function, and sets the prediction threshold when the loss of the training set converges. When the prediction accuracy rate of the test set is higher than the prediction threshold, the system parameter training is completed.

[0022] Further, the ventilator operation monitoring module real-time collects the ventilator operation data within the cycle, and sequentially inputs it into the safety margin standardization module, the initial dangerous point calibration module, the risk direction deviation detection module, the adaptive step size update module, and the iterative convergence determination module. If it is finally determined that the operation is abnormal, warning processing is performed.

[0023] The beneficial effects obtained by the present invention using the above scheme are as follows:

[0024] (1) Aiming at the problem that the general ventilator operation monitoring system is prone to missing the extreme value points in the transient state during the cycle, not responding promptly enough to the sudden risks of instantaneous airway obstruction and patient compliance mutation, generating false oscillation alarms, and thus resulting in poor operation monitoring accuracy, this scheme is gradient-oriented, continuously iteratively approaching the dangerous point, and accurately positioning the risk peak within the breathing cycle; Quantify the angle between the real risk trajectory and the ideal trajectory through risk direction deviation detection, and adjust the step size in real time to capture the transient nonlinear events of sudden airway obstruction; Thereby improving the operation monitoring accuracy.

[0025] (2)In view of the problems existing in the general ventilator operation monitoring system, such as slow mutation response, repeated oscillation in the non-linear local interval, missing the real change in the risk direction, prone to false alarms and missed alarms, and poor operation monitoring effect, this solution amplifies the angle normalization curve in the stable breathing interval to quickly approach the most dangerous point; performs linear convergence correction in the non-linear mutation interval, resists oscillation and divergence during severe non-linearity, and prevents missed alarms or false alarms; based on double convergence determination, it avoids false alarms caused by short-term noise, thereby improving the operation monitoring effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic flow chart of an intelligent ventilator operation monitoring system provided by the present invention.

[0027] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present invention.

[0029] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention.

[0030] Embodiment 1, referring to Figure 1 , an intelligent ventilator operation monitoring system provided by the present invention includes a data acquisition module, a safety margin evaluation function definition module, a safety margin standardization module, an initial dangerous point calibration module, a risk direction deviation detection module, an adaptive step size update module, an iterative convergence determination module, a system training module, and a ventilator operation monitoring module;

[0031] The data acquisition module collects the historical operation data of the ventilator, constructs an operation monitoring sample set; and sends the data to the safety margin evaluation function definition module;

[0032] The safety margin evaluation function definition module obtains the safety margin evaluation function by calculating the standardized margin of the deviation between each operation monitoring variable and its safety boundary, and sends the data to the safety margin standardization module;

[0033] The safety margin standardization module uses the empirical distribution function to map the original operation monitoring variables to the standard normal space, and sends the data to the initial dangerous point calibration module;

[0034] The initial dangerous point calibration module takes the most dangerous point within a cycle as the starting point to update the first iteration position, and sends the data to the risk direction deviation detection module;

[0035] The risk direction deviation detection module analyzes and detects the direction of the risk deviation, and sends the data to the adaptive step size update module;

[0036] The adaptive step size update module adjusts the step size based on interval division, and sends the data to the iteration convergence determination module;

[0037] The iteration convergence determination module performs iteration determination on the operation data within a cycle, and sends the data to the system training module;

[0038] The system training module trains the parameters within the system, and sends the data to the ventilator operation monitoring module;

[0039] The ventilator operation monitoring module monitors the real-time collected ventilator operation data within a cycle by connecting each system module in series.

[0040] Example 2, refer to Figure 1 , this example is based on the above example. The data acquisition module acquires the historical operation data of the ventilator, labels the operation status of the ventilator operation data within a cycle, and the operation status includes normal operation and abnormal operation. The historical operation data of the ventilator includes positive end-expiratory pressure, tidal volume, respiratory rate, end-tidal CO2 concentration, battery voltage, and oxygen pressure. The historical operation data of the ventilator is preprocessed to obtain an operation monitoring sample set. The preprocessing includes missing value and outlier processing, smoothing filtering, and vector conversion. A safety boundary is set for each parameter.

[0041] Example 3, refer to Figure 1 , this example is based on the above example. The safety margin evaluation function definition module defines the safety margin evaluation function , to ensure the safety of all indicators simultaneously, the worst margin is taken. For the i operation monitoring samples x i , the safety margin evaluation function is expressed as: ; where is the safety boundary of the jth monitoring variable; is the jth monitoring variable; is the standard deviation of the j-th monitoring variable; is the weight of the j-th monitoring variable, and the value range is [0, 1]; corresponds to the most dangerous dimension value inside the sample; for the operating monitoring sample set, calculate for each operating monitoring sample, and take the sample corresponding to the minimum value as the initial most dangerous point within one cycle.

[0042] Example 4, refer to Figure 1 , this example is based on the above example, and the safety margin normalization module performs a normalization transformation on each variable, expressed as: ; where is the standard normal inverse CDF; is the empirical distribution function of the i-th monitored quantity; obtain the mapped vector u; rewrite the safety margin evaluation function as: ; the vector u in each mapping space is mapped to the operating monitoring sample .

[0043] Example 5, refer to Figure 1 , this example is based on the above example, and the initial dangerous point calibration module in the standard normal space determines the initial most dangerous point during the respiratory cycle as , as the starting point of this cycle iteration, the first step size , position update: ; where is the position of the most dangerous point after the first iteration; is the value of the safety margin evaluation function at ; is the gradient; T is the transpose operation, and here the transpose of the gradient vector is taken.

[0044] Example 6, refer to Figure 1 , this example is based on the above example, and the risk direction deviation detection module calculates for the k-th iteration: ; ; ; where and are the positions of the most dangerous points at the k-th iteration and the (k - 1)-th iteration respectively; is the value of the safety margin evaluation function at ; is the ideal update vector; is the alignment degree; is the angle deviation; it quantifies the deviation of the iterative direction from the ideal direction in real time, reflecting the risks brought by the transient non-linearity of the ventilator-patient compliance mutation and airway obstruction; once a large turning angle is detected, the step size can be immediately adjusted to prevent iterative oscillation or divergence.

[0045] By performing the above operations, aiming at the problems existing in the general ventilator operation monitoring system, such as being prone to missing the extreme points of transients in the cycle, not being timely enough in reacting to the sudden risks of instantaneous airway obstruction and sudden changes in patient compliance, generating false oscillation alarms, and thus resulting in poor accuracy of operation monitoring, this solution is gradient-oriented, continuously iterates to approach the dangerous point, and precisely locates the risk peak within the respiratory cycle; quantifies the angle between the real risk trajectory and the ideal trajectory through risk direction deviation detection, adjusts the step size in real time, and captures the transient nonlinear events of sudden airway obstruction; thereby improving the accuracy of operation monitoring.

[0046] Embodiment Seven. Refer to Figure 1 , based on the above embodiment, the adaptive step size update module sets a corner threshold . For 's stable interval, adjusts the step size through a curve, expressed as: ; ; where t is the normalized angle deviation; is the step size; when the ventilator parameters fluctuate little, rapidly magnify the step size to accelerate finding the most dangerous point; greatly reduce the computational amount of each breath monitoring to meet the real-time requirement; for 's nonlinear mutation interval, corrects the step size through linear convergence, expressed as: ; ; ; ; where is the mutation auxiliary function; c is the balance coefficient; is the direction vector; is the smoothing term; take the minimum m such that ; ; where is the control parameter; T is the transpose operation, here taking the transpose of the gradient vector; in the case of drastic nonlinearities such as sudden airway blockage and sudden changes in patient compliance, ensure sufficient descent in each step; anti-oscillation and anti-divergence to prevent false alarms or missed alarms.

[0047] Embodiment Eight. Refer to Figure 1 , based on the above embodiment, for the stable interval, the iteration convergence determination module updates using ; Nonlinear mutation interval update: ; Convergence determination, avoiding local jitter in the nonlinearity and preventing missing drastic directional changes, expressed as: ; Calculate the reliability index , expressed as: ; ; where is the position of the most dangerous point at the (k + 1)-th iteration; is the momentum coefficient; and is the convergence threshold; a reliability index is obtained for each breathing cycle to quantify the safety margin; when occurs, it is determined that the operation is abnormal during the breathing cycle; otherwise, it is determined that the operation is normal; is the alarm probability threshold; is the cumulative distribution function of the standard normal distribution; for a stationary interval with little difference between the previous and next cycles, based on the momentum term, the position of the most dangerous point is found more quickly, reducing repeated iteration of minute noises.

[0048] By performing the above operations, aiming at the problems existing in the general ventilator operation monitoring system, such as slow mutation response, repeated oscillation in the non - linear local interval, missing real risk direction changes, being prone to false alarms and missed alarms, and poor operation monitoring effect, this solution magnifies the angular normalization curve in the breathing stable interval to quickly approach the most dangerous point; performs linear convergence correction in the non - linear mutation interval, resists oscillation and divergence during severe non - linearity, prevents missed alarms or false alarms; and avoids false alarms caused by short - term noises based on double - convergence determination, thereby improving the operation monitoring effect.

[0049] Embodiment Nine, refer to Figure 1 , based on the above - mentioned embodiment, the system training module divides the operation monitoring sample set into a test set and a training set, takes the safety boundaries of each monitoring variable, the weights of the monitoring variables, the corner threshold, the alarm probability threshold, the balance coefficient, the initial smoothing term, the control parameters, and the convergence threshold as system parameters, trains the system parameters based on the training set, uses the cross - entropy loss function, sets the prediction threshold when the loss for the training set converges, and when the prediction accuracy rate of the test set is higher than the prediction threshold, the system parameter training is completed.

[0050] Embodiment Ten, refer to Figure 1 , based on the above - mentioned embodiment, the ventilator operation monitoring module real - time collects the ventilator operation data within the cycle, and sequentially inputs it to the safety margin standardization module, the initial dangerous point calibration module, the risk direction deviation detection module, the adaptive step - size update module, and the iterative convergence determination module. If it is finally determined that the operation is abnormal, early warning processing is performed.

[0051] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0052] The above describes the present invention and its embodiments. Such description is not restrictive. What is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. In summary, if those of ordinary skill in the art are inspired by it and, without departing from the spirit of the present invention, design similar structural forms and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. An intelligent ventilator operation monitoring system, characterized in that: The system includes a data acquisition module, a safety margin evaluation function definition module, a safety margin standardization module, an initial dangerous point calibration module, a risk direction deviation detection module, an adaptive step size update module, an iterative convergence determination module, a system training module, and a ventilator operation monitoring module; The data acquisition module collects the historical operation data of the ventilator and constructs an operation monitoring sample set; The safety margin evaluation function definition module obtains a safety margin evaluation function by calculating the standardized margin of the deviation between each operation monitoring variable and its safety boundary; The safety margin standardization module maps the original operation monitoring variables to the standard normal space by using the empirical distribution function; The initial dangerous point calibration module uses the most dangerous point within a period as the starting point to update the first iteration position; The risk direction deviation detection module analyzes and detects the direction of the risk deviation; The adaptive step size update module adjusts the step size based on interval division; The iterative convergence determination module performs iterative determination on the operation data within a period; The system training module trains the parameters within the system; The ventilator operation monitoring module monitors the real-time collected operation data of the ventilator within a period by connecting each system module in series.

2. The intelligent ventilator operation monitoring system according to claim 1, wherein: The safety margin evaluation function definition module defines a safety margin evaluation function , to ensure the safety of all indicators simultaneously, the worst margin is taken. For the i operating monitoring samples x i , the safety margin evaluation function is expressed as: ; where is the safety boundary of the j-th monitoring variable; is the j-th monitoring variable; is the standard deviation of the j-th monitoring variable; is the weight of the j-th monitoring variable; corresponds to the most dangerous dimension value within the sample; for the operating monitoring sample set, calculate the of each operating monitoring sample, and take the sample corresponding to the minimum value as the initial most dangerous point within a cycle.

3. The intelligent ventilator operation monitoring system according to claim 2, wherein: The safety margin standardization module performs a normalization transformation on each variable, expressed as: ; where is the standard normal inverse CDF; is the empirical distribution function of the i-th monitored quantity; obtaining the mapped vector u; rewriting the safety margin evaluation function as: ; The vector u in each mapping space is mapped to the operation monitoring sample .

4. The intelligent ventilator operation monitoring system according to claim 3, wherein: The initial dangerous point calibration module determines the initial most dangerous point within the respiratory cycle in the standard normal space as , which serves as the starting point for the iteration of this cycle. The initial step size , and the position update is as follows: ; where is the position of the most dangerous point after the first iteration; is the value of the safety margin evaluation function at ; is the gradient; T is the transpose operation, and here the transpose of the gradient vector is taken.

5. The intelligent ventilator operation monitoring system according to claim 4, wherein: The risk direction deviation detection module performs the k-th iteration calculation: ; ; ; where and are the positions of the most dangerous points at the k-th iteration and the (k - 1)-th iteration, respectively; is the value of the safety margin evaluation function at ; is the ideal update vector; is the alignment degree; is the angular deviation.

6. The intelligent ventilator operation monitoring system according to claim 5, wherein: The adaptive step size update module sets a corner threshold , for the stable interval of, adjust the step size through a curve, expressed as: ; ; where t is the normalized angle deviation; is the step size; for the non-linear mutation interval of, correct the step size through linear convergence, expressed as: ; ; ; ; where, is the mutation auxiliary function; c is the balance coefficient; is the direction vector; is the smoothing term; take the minimum m such that ; ; where, is the control parameter; T is the transpose operation, here the transpose of the gradient vector is taken.

7. An intelligent ventilator operation monitoring system according to claim 6, characterized in that: For the stationary interval, the iteration convergence determination module updates and uses ; Nonlinear mutation interval update: ; Convergence determination, expressed as: ; Calculate the reliability index , expressed as: ; ; Among them, is the position of the most dangerous point at the (k + 1)-th iteration; is the momentum coefficient; and are the convergence thresholds; A reliability index is obtained for each breathing cycle to quantify the safety margin; When , it is determined that the operation is abnormal within the breathing cycle; Otherwise, it is determined that the operation is normal; is the alarm probability threshold; Cumulative distribution function of the standard normal distribution.

8. An intelligent ventilator operation monitoring system according to claim 7, characterized in that: The data acquisition module collects the historical operation data of the ventilator; labels the operation status of the operation data of the ventilator within a period, and the operation status includes normal operation and abnormal operation; preprocesses the historical operation data of the ventilator to obtain an operation monitoring sample set; sets a safety boundary for each parameter.

9. An intelligent ventilator operation monitoring system according to claim 8, wherein: The system training module divides the operation monitoring sample set into a test set and a training set, trains the system parameters based on the training set, uses the cross-entropy loss function, sets a prediction threshold when the loss of the training set converges, and the system parameter training is completed when the prediction accuracy rate of the test set is higher than the prediction threshold.

10. The intelligent ventilator operation monitoring system according to claim 9, characterized in that: The ventilator operation monitoring module collects the operation data of the ventilator within a period in real time, and sequentially inputs them into the safety margin standardization module, the initial dangerous point calibration module, the risk direction deviation detection module, the adaptive step size update module, and the iterative convergence determination module. If it is finally determined to be abnormal operation, warning processing is performed.

Images